U.S. patent application number 17/696099 was filed with the patent office on 2022-09-29 for posture recognition system, posture recognition method, and recording medium.
The applicant listed for this patent is Hitachi, Ltd.. Invention is credited to Takayuki AKIYAMA, Gabriele BLESER, Michael LORENZ, Takehiro NIIKURA, Didier STRICKER, Bertram TAETZ.
Application Number | 20220304593 17/696099 |
Document ID | / |
Family ID | 1000006273257 |
Filed Date | 2022-09-29 |
United States Patent
Application |
20220304593 |
Kind Code |
A1 |
NIIKURA; Takehiro ; et
al. |
September 29, 2022 |
POSTURE RECOGNITION SYSTEM, POSTURE RECOGNITION METHOD, AND
RECORDING MEDIUM
Abstract
Provided is a posture recognition system that can properly
respond to an error that is changed according to a user's motion. A
posture estimation section calculates posture data indicating a
posture of a user wearing clothing, on the basis of motion data
measured by a posture sensor attached to the clothing. An error
estimation section calculates error data which is an estimate of an
error that is occurring in the posture data, on the basis of the
motion data and the posture data.
Inventors: |
NIIKURA; Takehiro; (Tokyo,
JP) ; AKIYAMA; Takayuki; (Tokyo, JP) ; TAETZ;
Bertram; (Rheinland-Pfalz, DE) ; BLESER;
Gabriele; (Rheinland-Pfalz, DE) ; LORENZ;
Michael; (Rheinland-Pfalz, DE) ; STRICKER;
Didier; (Rheinland-Pfalz, DE) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Hitachi, Ltd. |
Tokyo |
|
JP |
|
|
Family ID: |
1000006273257 |
Appl. No.: |
17/696099 |
Filed: |
March 16, 2022 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
A61B 5/6804 20130101;
A61B 5/0004 20130101; A61B 2562/04 20130101; A61B 5/1116
20130101 |
International
Class: |
A61B 5/11 20060101
A61B005/11; A61B 5/00 20060101 A61B005/00 |
Foreign Application Data
Date |
Code |
Application Number |
Mar 24, 2021 |
JP |
2021-050639 |
Claims
1. A posture recognition system comprising: a posture estimation
section calculating, on a basis of sensor data measured by a
posture sensor attached to clothing, posture data indicating a
posture of a user wearing the clothing; and an error estimation
section calculating error data which is an estimate of an error
that is occurring in the posture data, on a basis of the sensor
data and the posture data.
2. The posture recognition system of claim 1, wherein the posture
estimation section generates, on a basis of sensor data measured by
each of a plurality of the posture sensors, the posture data
indicating the same posture of the user for each of the posture
sensors, and the error estimation section generates the error data
for each of the posture sensors.
3. The posture recognition system of claim 2, wherein the plurality
of posture sensors are attached to the clothing in such a manner as
to be arranged on a plurality of different body parts of the
user.
4. The posture recognition system of claim 2, wherein the plurality
of posture sensors are attached to the clothing in such a manner as
to be arranged on the same body part of the user.
5. The posture recognition system of claim 2, wherein the plurality
of posture sensors are attached to corresponding positions of the
plurality of pieces of clothing.
6. The posture recognition system of claim 1, wherein the posture
data and the error data are chronological data.
7. The posture recognition system of claim 1 further comprising: an
information generation section generating additional information
which is an evaluation of the error data.
8. A posture recognition method by a posture recognition system,
the posture recognition method comprising: calculating, on a basis
of sensor data measured by a posture sensor attached to clothing,
posture data indicating a posture of a user wearing the clothing;
and calculating error data which is an estimate of an error that is
occurring in the posture data, on a basis of the sensor data and
the posture data.
9. A non-transitory and tangible computer-readable recording medium
in which a program to be executed by a computer, the program
causing a computer to perform: calculating, on a basis of sensor
data measured by a posture sensor attached to clothing, posture
data indicating a posture of a user wearing the clothing; and
calculating error data which is an estimate of an error that is
occurring in the posture data, on a basis of the sensor data and
the posture data.
Description
CROSS-REFERENCE TO RELATED APPLICATION
[0001] The present application claims priority from Japanese
application JP2021-050639 filed on Mar. 24, 2021, the contents of
which is hereby incorporated by reference into this
application.
BACKGROUND OF THE INVENTION
1. Field of the Invention
[0002] The present disclosure relates to a posture recognition
system, a posture recognition method, and a recording medium.
2. Description of the Related Art
[0003] A posture recognition system that recognizes a user's
posture has drawn attention in recent years. A posture recognition
system is used, for example, to find burdens on workers engaged in
industrial or agricultural activities, ensure safety of the
workers, or improve work efficiency, and can send an instruction
according to a worker's posture to the worker.
[0004] A motion capture system that photographs a user by using
photographing equipment such as camera and a close-contact type
wearable sensor system that keeps a sensor such as an acceleration
sensor capable of detecting a worker's motion in close contact with
a user's body are known as posture recognition systems. With the
motion capture system, however, the user must remain within a
photographing range of the photographing equipment, which has put a
limitation on a user's action. With the close-contact type wearable
sensor system, there are also problems one example of which is that
the sensor itself has been a limitation on a worker's action.
[0005] In contrast, JP-2009-18158-A discloses a technique that
attaches a plurality of sensors such as temperature sensors and a
plurality of pieces of wireless equipment to clothing and
recognizes the user's posture from signals obtained by the
respective sensors. This technique makes it possible to reduce
limitations on the user's action.
[0006] In order to solve a problem of a posture recognition error
that occurs in a case where a sensor is changed in its position due
to slipping of clothing in a clothing type sensor system in which
the sensor is attached to clothing, a technique is disclosed in
"Shinya Namikawa, Yu Enokibori, Kenji Mase, "A study of preventive
measures against misalignment of clothing type sensor by arranging
a plurality of sensors of the same type in proximity," Multimedia,
Distributed, Cooperative, and Mobile (DICOMO2016) Symposium, July
2016, p. 1183-1189" in which a plurality of sensors are arranged in
proximity and an optimal sensor is adopted from among these
sensors.
[0007] In the technique described in JP-2009-18158-A, no
consideration has been given to an error that occurs in the case
where a sensor is attached to clothing. Although consideration has
been given to an error that occurs when a sensor is changed in its
position due to slipping of clothing in the technique described in
"Shinya Namikawa, Yu Enokibori, Kenji Mase, "A study of preventive
measures against misalignment of clothing type sensor by arranging
a plurality of sensors of the same type in proximity," Multimedia,
Distributed, Cooperative, and Mobile (DICOMO2016) Symposium, July
2016, p. 1183-1189," what is done here is merely to select, from
among the plurality of sensors arranged in proximity, the sensor
that provides the smallest error that occurs in the case where a
specific motion is made. Accordingly, it has been difficult to
properly respond to an error whose cause and magnitude are changed
according to the user's motion.
[0008] It is an object of the present disclosure to provide a
posture recognition system, a posture recognition method, and a
recording medium that can properly respond to an error that is
changed according to a user's motion.
SUMMARY OF THE INVENTION
[0009] A posture recognition system according to an aspect of the
present disclosure includes a posture estimation section and an
error estimation section. The posture estimation section
calculates, on the basis of sensor data measured by a posture
sensor attached to clothing, posture data indicating a posture of a
user wearing the clothing. The error estimation section calculates
error data, which is an estimate of an error that is occurring in
the posture data, on the basis of the sensor data and the posture
data.
[0010] According to the present invention, it is possible to
properly respond to an error that is changed according to a user's
motion.
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] FIG. 1 is a block diagram illustrating a posture recognition
system according to embodiment 1 of the present disclosure;
[0012] FIG. 2 is a schematic diagram illustrating an example in
which a user wears clothing to which sensor sections are
attached;
[0013] FIG. 3 is a conceptual diagram for describing an example of
a function of a posture estimation section;
[0014] FIGS. 4A to 4D are diagrams for describing examples of
errors that occur in an estimated posture;
[0015] FIG. 5 is a diagram illustrating examples of classes into
which an error is classified;
[0016] FIG. 6 is a flowchart for describing operation of the
posture recognition system;
[0017] FIGS. 7A and 7B are diagrams for describing examples of
errors according to a posture sensor position; and
[0018] FIGS. 8A and 8B are diagrams for describing other examples
of errors according to the posture sensor position.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0019] A description will be given below of embodiments of the
present disclosure with reference to drawings. It should be noted,
however, that the following embodiments are illustrative for
description of the present disclosure and are not intended to limit
the scope of the present disclosure to these embodiments. A person
skilled in the art can carry out the present disclosure in various
other manners without departing from the scope of the present
disclosure.
[0020] In the configuration of the invention described below, the
same portions or portions having the same function may be denoted
by the same reference signs in common in different drawings for
omission of redundant description. In the case where there are a
plurality of elements having the same or similar functions, such
elements may be described by adding different suffixes to the same
reference sign. It should be noted, however, that in the case where
there is no need to distinguish between the plurality of elements,
such elements may be described by omitting suffixes.
[0021] Actual positions, sizes, shapes, and regions of the
components illustrated in the drawings may not be given to
facilitate the understanding of the present disclosure.
Accordingly, the present disclosure is not limited to the
positions, sizes, shapes, and regions disclosed in the drawings.
Even if a component is illustrated in singular form in the present
specification, there may be the plurality of components unless
otherwise indicated in context.
Embodiment 1
[0022] <1. Overall Configuration of the System>
[0023] FIG. 1 is a block diagram illustrating a posture recognition
system according to embodiment 1 of the present disclosure. A
posture recognition system 1 illustrated in FIG. 1 estimates a
posture of a user 210 and includes a sensor section 110 and an
information processing apparatus 120. The sensor section 110 and
the information processing apparatus 120 are connected to each
other in a wired or wireless manner for communication. It should be
noted that the sensor section 110 and the information processing
apparatus 120 may be connected to each other via a communication
network.
[0024] The sensor section 110 is a wearable unit attached to
clothing 111 of the user 210. The sensor section 110 may be, for
example, pasted to the clothing 111 by using adhesive tape or
embedded in advance in the clothing 111. The clothing 111 is, for
example, common clothing that is not in close contact with a body
of the user 210 such as work clothing.
[0025] The sensor section 110 includes a posture sensor 112 and a
communication section 113. The posture sensor 112 is a motion
measurement sensor for measuring the motions of the user 210. The
motions refer to motions in general including not only motions
associated with industrial and agricultural activities but also
motions for specific purposes such as dancing, gymnastics, and
performance with musical instruments.
[0026] FIG. 2 is a schematic diagram illustrating an example in
which the user 210 is wearing the clothing 111 to which the sensor
section 110 is attached.
[0027] In FIG. 2, the clothing 111 has the plurality of posture
sensors 112 as the sensor sections 110 and a sensor hub 220 that
collects motion data 140 (refer to FIG. 1) which is sensor data
measured by the plurality of posture sensors 112. Each of the
posture sensors 112 is connected to the sensor hub 220 directly or
by way of other posture sensor 112 and sends the motion data 140 to
the sensor hub 220. It should be noted that the posture sensors 112
and the sensor hub 220 may be connected to each other in a wireless
or wired manner.
[0028] Although the posture sensors 112 are attached to the
clothing 111 in such a manner as to be arranged on respective body
parts of the user 210 in the example illustrated in FIG. 2, the
arrangement of the posture sensors 112 is not limited to this
example. For example, the posture sensors 112 may be attached to
the clothing 111 in such a manner as to be arranged on only some
body parts of the user such as upper or lower body.
[0029] Only one type or a plurality of types of the posture sensors
112 may be used, and the type of the posture sensors 112 may be
selected as appropriate according to a target's posture to be
estimated. For example, in the case where the posture of the user
210 is directly estimated, it is preferred that the type of sensors
that can measure the positions or motions of the respective body
parts of the user 210 such as acceleration sensor or position
sensor be used. The type of the posture sensors 112 is not limited
to this example, and gyro sensor, geomagnetic sensor, myoelectric
sensor, acceleration sensor, video sensor, or audio sensor, for
example, may also be used.
[0030] The sensor hub 220 sends the motion data 140 from the
respective posture sensors 112 to the information processing
apparatus 120 via the communication section 113. Although not
illustrated in FIG. 2, the communication section 113 is
incorporated, for example, in the sensor hub 220. It may be
possible to feed power to the posture sensors 112 and the sensor
hub 220 from a power feeding apparatus (not illustrated) such as
battery.
[0031] Referring back to the description of FIG. 1, the information
processing apparatus 120 functions as a posture estimation
apparatus that estimates the posture of the user 210 on the basis
of the motion data 140 from the sensor section 110. The information
processing apparatus 120 can include a common server. For example,
the information processing apparatus 120 can include a server that
includes an input apparatus, an output apparatus, a processing
apparatus, and a storage apparatus as hardware. In the present
embodiment, a program stored in the storage apparatus is read and
executed by the processing apparatus, as a result of which a
variety of functions are realized due to coordination between the
processing apparatus and other hardware as necessary.
[0032] It should be noted that the information processing apparatus
120 may include a single server or other computer system in which
optional portions of the input apparatus, the output apparatus, the
processing apparatus, and the storage apparatus are connected to
each other by a network. Functions equivalent to those realized by
the above program may be realized by hardware such as FPGA (Field
Programmable Gate Array) or ASIC (Application Specific Integrated
Circuit). A neural network which will be described later may be
realized by FPGA.
[0033] The information processing apparatus 120 includes a
communication section 121, a posture estimation section 122, an
error estimation section 123, an information generation section
124, and a control section 125 as functional components.
[0034] The communication section 121 communicates with external
apparatuses such as the sensor section 110. For example, the
communication section 121 receives the motion data 140 from the
communication section 113 of the sensor section 110. The
communication section 121 sends posture information 150 indicating
the posture of the user 210 estimated on the basis of the motion
data 140. Although destinations of the posture information 150 are
not specifically limited, the posture information 150 is sent, for
example, to the storage apparatus that stores the posture
information 150, a display apparatus that displays the posture
information 150, or an analysis apparatus that analyzes the posture
information 150.
[0035] The posture estimation section 122 estimates the posture of
the user 210 on the basis of the motion data 140 received by the
communication section 121 and calculates posture data indicating
the estimated posture which is an estimate of the posture. The
error estimation section 123 estimates an error that is occurring
in the posture data calculated by the posture estimation section
122 and calculates error data indicating the estimated error. The
posture estimation section 122 and the error estimation section 123
can be configured, for example, by using a neural network.
[0036] The information generation section 124 generates, on the
basis of the posture data calculated by the posture estimation
section 122 and the error data calculated by the error estimation
section 123, additional information which is an evaluation of the
error data. The control section 125 controls the information
processing apparatus 120 as a whole.
[0037] <2. Posture Estimation Section>
[0038] FIG. 3 is a conceptual diagram for describing an example of
a function of the posture estimation section 122. The posture
estimation section 122 calculates posture data indicating the
estimated posture of the user 210 as described above. The estimated
posture specifically refers to an estimate of the posture or
postures of one or a plurality of given body parts of the user 210.
The body part refers to a body part that can be considered as a
single rigid body (e.g., upper arm, forearm, or shin).
[0039] In the present embodiment, the posture estimation section
122 calculates a feature quantity representing a feature of the
estimated posture as posture data. The feature quantities
correspond to states of moving parts (joints) connected to body
parts and are, for example, elbow bending angle, lower back bending
angle, head orientation, and opening leg angle.
[0040] The type and arrangement of the posture sensors 112 are
selected according to the target's posture to be estimated. A
description will be given below of an example in which a lower back
bending angle is calculated as a feature quantity which is posture
data by using an acceleration sensor as the posture sensor 112
unless otherwise specified. In this case, the posture estimation
section 122 includes a lower back state estimation section 301 that
calculates a lower back bending angle as illustrated in FIG. 3.
[0041] The lower back state estimation section 301 calculates, for
example, a rotation angle of the lower back in a fore-and-aft
direction as a bending angle of the lower back which is a feature
quantity on the basis of the motion data 140 from the four posture
sensors 112 arranged on shoulders, a chest, and the lower back of
the user 210 illustrated in FIG. 3. Since values of the motion data
140 are changed over time, the lower back state estimation section
301 acquires the lower back bending angle as chronological data
302. The acquired chronological data 302 may be stored, for
example, in the storage apparatus within the information processing
apparatus 120 or in an external storage apparatus.
[0042] The lower back state estimation section 301 includes, for
example, a learning model built by known supervised learning by
using a deep neural network (hereinafter referred to as a DNN) that
has the motion data 140 as an explanatory variable and a feature
quantity as an objective variable. It should be noted that the
lower back state estimation section 301 may calculate a feature
quantity from the motion data 140 of the posture sensors 112
without using a DNN. For example, the lower back state estimation
section 301 can estimate the lower back bending angle by acquiring,
in advance, the motion data 140 (acceleration) in a given posture
(e.g., upright posture) of the user as an initial state or by
acquiring the motion data 140 of other sensor such as position
sensor different from the acceleration sensor without using a
DNN.
[0043] It should be noted that in the case where a feature quantity
other than the lower back bending angle is calculated, it is only
necessary to add, in keeping with the feature quantity, a
functional section such as elbow rotation angle estimation section
or head orientation estimation section which calculates the feature
quantity to the posture estimation section 122.
[0044] <3. Error Estimation Section>
[0045] FIGS. 4A to 4D are diagrams for describing examples of
errors that occur in posture data (feature quantity) calculated by
the posture estimation section 122 due to deformation of the
clothing 111.
[0046] Normally, the posture sensors 112 are attached to the
clothing 111 such that the positions and orientations thereof with
respect to the body of the user 210 are as expected in advance as
illustrated in FIG. 4A. However, in the case where the user 210
makes some kind of motion, or in the case where the clothing 111
does not partially fit a body shape of the user 210, the positions
and orientations of the posture sensors 112 may deviate from the
expected positions and orientations.
[0047] For example, in the case where there is a slight slack in
the clothing 111 on an abdomen as illustrated in FIG. 4B, there is
a possibility that a minor error may occur due to slight deviation
in the orientation of the posture sensor 112a which is the posture
sensor 112 arranged on the abdomen. In the case where the user 210
extends his or her arm upward as illustrated in FIG. 4C, a hem of
the clothing 111 is raised in keeping with the movement of the arm,
which may cause the position of the posture sensor 112a arranged on
the abdomen to deviate and result in a medium error. In the case
where the user tilts forward as illustrated in FIG. 4D, the posture
sensor 112b which is the posture sensor 112 arranged on the chest
tilts in keeping with the forward tilting of the user 210 if the
posture sensor 112b is in close contact with the body of the user
210. However, in the case where the posture sensor 112b is attached
to the clothing 111, the posture sensor 112b may remain
approximately vertical relative to the ground due to deformation of
the chest of the clothing 111 (e.g., slack). In this case, there is
a possibility that a large error may occur because no information
can be acquired that indicates that the user 210 is tilting forward
from the motion data 140 of the posture sensor 112b.
[0048] In the case where the posture sensors 112 are attached to
the clothing 111 as described above, an error may occur in posture
data calculated by the posture estimation section 122 due to
deformation of the clothing 111. The error estimation section 123
calculates error data which is an estimate of an error that is
occurring in the posture data on the basis of the motion data 140
and the posture data. It should be noted that since error data is
changed over time, the error estimation section 123 acquires an
error as chronological data. The acquired chronological data may be
stored, for example, in the storage apparatus within the
information processing apparatus 120 or in an external storage
apparatus.
[0049] The error estimation section 123 includes, for example, a
learning model built by known supervised learning by using a DNN
that has the motion data 140 and posture data as explanatory
variables and an error as an objective variable. A learning model
can be built, for example, by causing the user 210 wearing the
clothing 111 to which the posture sensors 112 are attached to
perform various tasks, acquiring an error between a user's actual
posture and an estimated posture (posture data), and using known
supervised learning in which the motion data 140, posture data, and
error are used as training data.
[0050] It should be noted that the motion data 140 used as training
data can be selected as appropriate according to the target's
posture to be estimated and the body part. For example, in the case
where the lower back bending angle is estimated, the motion data
140 used as training data may be motion data from the posture
sensors 112 arranged on the abdomen and chest or data obtained by
adding motion data from the posture sensors 112 arranged on the
shoulders and upper arms to the motion data from the posture
sensors arranged on the abdomen and chest. In the case where the
lower back bending angle is estimated, posture data used as
training data may be the lower back bending angle or data obtained
by adding feature quantities associated with the upper arms to the
lower back bending angle.
[0051] As means of acquiring the user's actual posture, the actual
posture may be visually determined from a video image acquired by
using imaging equipment such as camera. Alternatively, a known
motion capture system or a close-contact type wearable sensor
system may be used.
[0052] The error estimation section 123 may estimate an error by
classifying the error into a plurality of classes. FIG. 5 is a
diagram illustrating examples of classes into which an error is
classified. In the example illustrated in FIG. 5, an error is
classified into four classes from error level 0 to error level 3.
Error level 0 is a class with almost no error. Error level 1 is a
class with a minor error. Error level 2 is a class with a medium
error. Error level 3 is a class with a major error.
[0053] <4. Information Generation Section>
[0054] The information generation section 124 generates additional
information which is an evaluation of error data on the basis of
posture data calculated by the posture estimation section 122 and
error data calculated by the error estimation section 123 and
generates the posture information 150 that includes the posture
data, the error data, and the additional information.
[0055] Additional information is, for example, evaluation
information indicating availability or reliability of the estimated
posture or notification information according to an error. The
evaluation information refers specifically to information
indicating the extent to which an estimated posture is reliable and
may be calculated by using an error itself or by using a
statistical value such as time average of an error. Notification
information is, for example, information for notifying the user 210
or an administrator of the posture recognition system 1 that an
error is large in the case where the error or the statistical value
thereof is equal to or larger than a threshold. Notification
information may be, for example, output as an alert tone from the
information processing apparatus held by the user 210 or the
administrator of the posture recognition system 1.
[0056] <5. Operation>
[0057] FIG. 6 is a flowchart for describing operation of the
posture recognition system 1.
[0058] First, each of the posture sensors 112 of the sensor section
110 regularly or continuously measures the motion data 140 and
outputs the motion data 140 to the sensor hub 220. The sensor hub
220 sends the motion data 140 sent from each of the posture sensors
112 to the information processing apparatus 120 via the
communication section 113 (step S101).
[0059] When the communication section 121 of the information
processing apparatus 120 receives the motion data 140, the posture
estimation section 122 estimates the posture of the user 210 on the
basis of the motion data 140 and calculates a feature quantity of
the estimated posture as posture data indicating the estimated
posture (step S102).
[0060] The error estimation section 123 estimates an error that is
occurring in the feature quantity calculated by the posture
estimation section 122 on the basis of the motion data 140 and the
feature quantity and calculates error data that indicates the
estimated error (step S103).
[0061] The information generation section 124 generates additional
information which is an evaluation of the error data on the basis
of the feature quantity and the error data (step S104).
[0062] The information generation section 124 generates the posture
information 150 that includes the feature quantity, the error data,
and the additional information. The communication section 121 sends
the posture information (step 5105) and terminates the process.
[0063] <6. Advantageous Effect of the Embodiments>
[0064] According to the embodiment described above, the posture
estimation section 122 calculates posture data indicating the
posture of the user 210 wearing the clothing 111 on the basis of
the motion data 140 measured by the posture sensors 112 attached to
the clothing 111. The error estimation section 123 calculates error
data which is an estimate of an error that is occurring in the
posture data on the basis of the motion data 140 and the posture
data. Accordingly, it becomes possible to recognize the cause and
magnitude of the error that are changed according to the motion of
the user 112 due to attachment of the posture sensors 112 to the
clothing 111, which makes it possible to properly respond to the
error. For example, it becomes possible for an application program
that performs a variety of processes on the basis of posture data
to perform the processes in consideration of the error, which makes
it possible to perform the processes on the basis of the posture
data with more accuracy.
[0065] In the present embodiment, posture data and error data are
chronological data. Accordingly, it becomes possible to find the
change in error over time, which makes it possible to more properly
respond to the error.
[0066] In the present embodiment, the information generation
section 124 generates additional information which is an evaluation
of error data. Accordingly, it becomes possible to issue
notification about reliability of posture data, which makes it
possible to properly respond to an error with ease.
Embodiment 2
[0067] In the present embodiment, a description will be given of an
example of the posture recognition system 1 that can reduce an
error by generating posture data and error data corresponding to
each of the plurality of posture sensors 112 attached to different
positions. The posture sensors 112 are specifically attached to the
clothing 111 in such a manner as to be arranged on a plurality of
different body parts of the user 210.
[0068] FIGS. 7A and 7B are diagrams for describing a change in
error according to the position of the posture sensor 112.
[0069] FIG. 7A illustrates an example in which the posture sensor
112b is arranged on the chest as in embodiment 1 (refer to FIG.
4D). In this case, when the user 210 tilts forward, the posture
sensor 112b needs to tilt so as to accurately estimate the posture
as the user 210 tilts forward. However, since the posture sensor
112b is attached to the clothing 111, the posture sensor 112b may
become approximately vertical relative to the ground due to
deformation of the chest of the clothing 111. In this case, there
is a possibility that a large error may occur because no
information can be acquired that indicates that the user 210 is
tilting forward from the motion data 140 of the posture sensor
112b.
[0070] FIG. 7B illustrates an example in which the posture sensor
is arranged on an upper back rather than on the chest. In this
case, when the user 210 tilts forward, a posture sensor 112c which
is the posture sensor 112 arranged on the upper back tilts forward
as the user 210 tilts forward, which makes it unlikely that a large
error will occur in the motion data 140 of the posture sensor
112c.
[0071] In the present embodiment, the posture estimation section
122 calculates posture data indicating the same posture of the user
210 for each posture sensor 112 on the basis of the motion data 140
measured by each of the plurality of posture sensors 112 (e.g., the
posture sensors 112b and 112c illustrated in FIGS. 7A and 7B)
arranged on the plurality of different body parts of the user 210.
The error estimation section 123 calculates, for each posture
sensor 112, error data indicating an error that is occurring in
that posture sensor 112. The information generation section 124 may
generate additional information for each posture sensor 112 or may
generate, as additional information, information obtained by
comparing pieces of error data corresponding to the respective
posture sensors 112 with each other.
[0072] It should be noted that although the plurality of posture
sensors 112 are attached to different positions in the present
embodiment, the posture sensors 112 may be attached in different
manners in addition to or in place of the positions where the
posture sensors 112 are attached. The manners in which the posture
sensors 112 are attached include, for example, pasting to the
clothing 111 with adhesive tape and embedding in the clothing 111
in advance.
[0073] According to the embodiment described above, it is possible
to confirm how error data is changed according to the positions
where the posture sensors 112 are attached and the manners in which
the posture sensors 112 are attached. Accordingly, it is possible
to find the position and manner in which the posture sensors 112
are attached that provide a small error, which makes it possible to
realize the arrangement of the posture sensors 112 with a smaller
error.
Embodiment 3
[0074] In the present embodiment, a description will be given of an
example of the posture recognition system 1 that can reduce an
error by generating posture data and error data corresponding to
each of the plurality of posture sensors 112 attached to different
pieces of clothing. The posture sensors 112 are specifically
attached to corresponding positions of respective pieces of the
clothing 111 of the same type and different sizes. We assume here
that the sizes are S, M, and L in the order from small to large. It
should be noted, however, that the posture sensors 112 may be
attached to different types of clothing.
[0075] The posture estimation section 122 calculates posture data
indicating the same posture of the user 210 on the basis of the
motion data 140 measured by each of the posture sensors 112
attached to the plurality of pieces of clothing 111, respectively,
for each posture sensor 112 (i.e., for each piece of the clothing
111). Here, the same user 210 puts on the plurality of pieces of
clothing 111 of different sizes in sequence and makes a given
motion, and the posture estimation section 122 calculates feature
quantities in sequence by calculating posture data corresponding to
the respective pieces of the clothing 111 in sequence. It should be
noted, however, that the posture estimation section 122 may
calculate posture data in parallel for the plurality of pieces of
clothing 111 put on by the different users 210. A single motion or
a series of motions such as radio calisthenics may be made as a
given motion.
[0076] The error estimation section 123 calculates, for each
posture sensor 112, error data indicating an error that is
occurring in that posture sensor 112. The information generation
section 124 may generate additional information for each posture
sensor 112 or may generate, as additional information, information
obtained by comparing pieces of error data corresponding to the
respective posture sensors 112 with each other.
[0077] According to the embodiment described above, it is possible
to generate error data for each of the plurality of pieces of
clothing. Accordingly, it becomes possible to select the piece of
clothing 111 that suits the user 210 from among the plurality of
pieces of clothing 111 such as selecting the piece of clothing 111
that provides the smallest error for the body shape of the user
120.
Embodiment 4
[0078] In the present embodiment, a description will be given of an
example of the posture recognition system 1 that can reduce an
error by generating posture data and error data corresponding to
each of the plurality of posture sensors 112 attached to different
positions as in embodiment 2. It should be noted, however, that, in
the present embodiment, an example will be described in which the
posture sensors 112 are attached to the clothing 111 in such a
manner as to be arranged on the same body part of the user 210.
[0079] FIGS. 8A and 8B are diagrams illustrating states in which
the posture sensors 112 are attached to the clothing 111 in such a
manner as to be arranged on the same body part of the user 210. The
examples in FIGS. 8A and 8B illustrate states in which the user 210
is wearing the clothing 111 to which three posture sensors 112d to
112f as the posture sensors 112 are attached at positions
corresponding to the right forearm and in which the user has his or
her arm raised.
[0080] In the example in FIG. 8A, even when the user 210 raises his
or her arm, all the posture sensors 112d to 112f remain on the
forearm. In this case, the values of error data based on the motion
data 140 corresponding to the respective posture sensors 112d to
112f are equivalent. In contrast, in the example in FIG. 8B, as the
user 210 raises his or her arm, a sleeve of the clothing 111 slides
down, which causes some of the posture sensors 112d to 112f
(posture sensor 112f here) arranged on the forearm to move to the
upper arm located more on the side of the shoulder than the
forearm. In this case, if error data is calculated on the basis of
the motion data 140 corresponding to each of the posture sensors
112d to 112f, the error data based on the motion data 140 of the
posture sensor 112f that moves to the upper arm is larger than the
error data based on the motion data 140 of the other posture
sensors, namely, the posture sensors 112d and 112e.
[0081] In the present embodiment, the posture estimation section
122 calculates posture data indicating the same posture of the user
210 for each posture sensor 112 on the basis of the motion data 140
measured by each of the plurality of posture sensors 112 (e.g., the
posture sensors 112d to 112f illustrated in FIGS. 8A and 8B)
arranged on the same body part of the user 210. The error
estimation section 123 calculates, for each posture sensor 112,
error data indicating an error that is occurring in that posture
sensor 112. The information generation section 124 may generate
additional information for each posture sensor 112 or may generate,
as additional information, information obtained by comparing pieces
of error data corresponding to the respective posture sensors 112
with each other.
[0082] It should be noted that although the plurality of posture
sensors 112 are attached to the positions corresponding to the
forearm in the present embodiment, the body part to which the
posture sensors are attached is not limited to the forearm, and the
posture sensors may be attached to a body part that can be
considered a single rigid body (e.g., upper arm, forearm,
shin).
[0083] As described above, it is possible, in the present
embodiment, to realize more accurate posture recognition such as
recognizing in what manner the clothing 111 is deformed.
[0084] The posture recognition system 1 in each of the embodiments
described above is applicable, for example, to a work support
system or an education system. The work support system can be used,
for example, for training of maintenance tasks. The education
system can be used, for example, for practicing dancing or yoga
poses.
* * * * *